Line data    Source code

       1              : /**
       2              :  * Copyright (c) 2025 Huawei Technologies Co., Ltd.
       3              :  * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
       4              :  * CANN Open Software License Agreement Version 2.0 (the "License").
       5              :  * Please refer to the License for details. You may not use this file except in compliance with the License.
       6              :  * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
       7              :  * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
       8              :  * See LICENSE in the root of the software repository for the full text of the License.
       9              :  */
      10              : 
      11              : /*!
      12              :  * \file kernel_vec_bilinearinterpolation_check.cpp
      13              :  * \brief
      14              :  */
      15              : 
      16              : #include "kernel_check_params.h"
      17              : #include "kernel_vec_bilinearinterpolation_check.h"
      18              : 
      19              : namespace AscendC {
      20              : namespace check {
      21              : 
      22              : const uint64_t ELE_PER_REPEAT = 8; // 1 repeat && repeatMode = 1 -> 8 src1 elements
      23              : 
      24            0 : bool TikcppVecBilinearInterpolationCheck::CheckTensorSize(std::vector<uint64_t> maskArray)
      25              : {
      26            0 :     uint8_t n = param_.hRepeat;
      27            0 :     uint8_t m = param_.vRepeat;
      28              :     // max element calculated in a repeat
      29            0 :     uint64_t maskVal = (maskArray.size() == 1) ? maskArray[0] : GetMaskLength(maskArray, param_.src0DtypeBytes);
      30              : 
      31              :     // src0OffsetLocal: total: m * n block, but last block is related to maskVal
      32            0 :     uint64_t eleNumPerOffset = ONE_BLK_SIZE / param_.src0DtypeBytes; // each offset element -> element in src0
      33              :     // example: maskVal = 17, dtype = half, thus offset needs at least 2 elements to pick blocks in src0
      34            0 :     uint64_t offsetNumPerRep = (maskVal + eleNumPerOffset - 1) / eleNumPerOffset;
      35            0 :     uint64_t expectedSize = (m * n - 1) * ONE_BLK_SIZE + offsetNumPerRep * param_.offsetDtypeBytes;
      36            0 :     ASCENDC_CHECK(CheckTensorSizeOverflow(expectedSize, param_.offsetSize, "src0OffsetLocal", "BilinearInterpolation"));
      37              : 
      38              :     // src0Local: needs to read values in src0OffsetLocal. Thus make optimistic check
      39              :     // example: maskVal = 7, only need 7 elements at most; maskVal = 16, probably src0Local all points to same block
      40            0 :     expectedSize = (maskVal <= eleNumPerOffset) ? maskVal * param_.src0DtypeBytes : ONE_BLK_SIZE;
      41            0 :     ASCENDC_CHECK(CheckTensorSizeOverflow(expectedSize, param_.src0Size, "src0Local", "BilinearInterpolation"));
      42              : 
      43              :     // src1Local: similar to src0Offset
      44            0 :     expectedSize = (param_.repeatMode) ? ELE_PER_REPEAT * (m * n - 1) + offsetNumPerRep : m * n; // in unit of elements
      45            0 :     ASCENDC_CHECK(CheckTensorSizeOverflow(
      46              :         expectedSize * param_.src1DtypeBytes, param_.src1Size, "src1Local", "BilinearInterpolation"));
      47              : 
      48              :     // dstLocal:
      49            0 :     uint64_t blockLen = ONE_BLK_SIZE / param_.dstDtypeBytes; // ele num in one block
      50            0 :     uint64_t blkNumLastRep = DivCeil(maskVal, blockLen);     // last repeat needs x blocks for maskLen elements
      51            0 :     uint64_t eleNumLastBlk = ((maskVal % blockLen) != 0) ? (maskVal % blockLen) : blockLen;
      52            0 :     uint64_t dstLastRepStart = (m - 1) * param_.vROffset;                                               // unit of elem
      53            0 :     uint64_t dstLastRepEleNum = (blkNumLastRep - 1) * param_.dstBlockStride * blockLen + eleNumLastBlk; // unit of elem
      54            0 :     expectedSize = (dstLastRepStart + dstLastRepEleNum) * param_.dstDtypeBytes;
      55            0 :     ASCENDC_CHECK(CheckTensorSizeOverflow(expectedSize, param_.dstSize, "dstLocal", "BilinearInterpolation"));
      56            0 :     return true;
      57              : }
      58              : 
      59            0 : bool TikcppVecBilinearInterpolationCheck::CheckAllLowLevel(std::vector<uint64_t> maskArray)
      60              : {
      61            0 :     uint32_t maxByteLen = std::max(std::max(param_.dstDtypeBytes, param_.src0DtypeBytes), param_.src1DtypeBytes);
      62            0 :     ASCENDC_CHECK(UpdateMaskArrayAndCheck(maskArray, maxByteLen));
      63              : 
      64            0 :     const std::string supportPos = "VECIN/VECOUT/VECCALC";
      65            0 :     ASCENDC_CHECK(
      66              :         CheckTensorScope(param_.dstLogicPos, static_cast<uint8_t>(HardWareIndex::UB), "dstLocal", supportPos));
      67            0 :     ASCENDC_CHECK(
      68              :         CheckTensorScope(param_.src0LogicPos, static_cast<uint8_t>(HardWareIndex::UB), "src0Local", supportPos));
      69            0 :     ASCENDC_CHECK(CheckTensorScope(
      70              :         param_.offsetLogicPos, static_cast<uint8_t>(HardWareIndex::UB), "src0OffsetLocal", supportPos));
      71            0 :     ASCENDC_CHECK(
      72              :         CheckTensorScope(param_.src1LogicPos, static_cast<uint8_t>(HardWareIndex::UB), "src1Local", supportPos));
      73              : 
      74            0 :     ASCENDC_CHECK(CheckTensorAddrAlign(param_.dstAddr, param_.dstPos, ONE_BLK_SIZE, "dstLocal"));
      75            0 :     ASCENDC_CHECK(CheckTensorAddrAlign(param_.src0Addr, param_.src0Pos, ONE_BLK_SIZE, "src0Local"));
      76            0 :     ASCENDC_CHECK(CheckTensorAddrAlign(param_.offsetAddr, param_.offsetPos, ONE_BLK_SIZE, "src0OffsetLocal"));
      77            0 :     ASCENDC_CHECK(CheckTensorAddrAlign(param_.src1Addr, param_.src1Pos, ONE_BLK_SIZE, "src1Local"));
      78              : 
      79            0 :     ASCENDC_CHECK(CheckBufferSizeOverFlow(
      80              :         param_.src0Size, GlobalParams::Instance().bufferSizeMap.at(param_.src0Pos),
      81              :         "Failed to check src0Local tensor buffersize"));
      82            0 :     ASCENDC_CHECK(CheckBufferSizeOverFlow(
      83              :         param_.src1Size, GlobalParams::Instance().bufferSizeMap.at(param_.src1Pos),
      84              :         "Failed to check src1Local tensor buffersize"));
      85            0 :     ASCENDC_CHECK(CheckBufferSizeOverFlow(
      86              :         param_.offsetSize, GlobalParams::Instance().bufferSizeMap.at(param_.offsetPos),
      87              :         "Failed to check src0OffsetLocal tensor buffersize"));
      88            0 :     ASCENDC_CHECK(CheckBufferSizeOverFlow(
      89              :         param_.dstSize, GlobalParams::Instance().bufferSizeMap.at(param_.dstPos),
      90              :         "Failed to check dstLocal tensor buffersize"));
      91              : 
      92            0 :     ASCENDC_CHECK(CheckTensorSize(maskArray));
      93            0 :     return true;
      94            0 : }
      95              : 
      96              : } // namespace check
      97              : } // namespace AscendC